Automated Team Analysis

We have created an agent for analyzing and improving synthetic teams. The agent is built in a bottom-up fashion using little specific domain knowledge. In lieu of extensive domain knowledge, data mining and inductive learning techniques are used in an attempt to isolate the key issues determining the successes or failures of these teams. This approach has been applied to the RoboCup domain, with a current focus on analyzing shots on goal and with future plans for assists, passing, and general teamwork.

Authors

• Taylor Raines
• Milind Tambe Harvard University Google Research
• Stacy Marsella
• University of Southern California

Keywords

No keywords are indexed for this paper.